This thesis presents a framework for the automatic identification of the state of an oil drilling system from sensor data. The reliable detection of states is a prerequisite for the identification of operations. Although the framework has been developed for monitoring drilling, it is generally applicable to data fusion models for the generation of features and decision making. The system identifies specific states of the equipment and/or process dependent on predefined sensor information extracted dynamically from the sensor data. Three fundamental types of states are defined: Cluster States, Trend States, and Shape States. Cluster States are defined by discriminating data into clusters using ``Expectation Maximization'', ``Envelope" and ``Otsu" algorithms. Trend States are detected in sensor measurements by applying Piecewise Linear Approximation algorithm where the final trend states are determined after a number of merging operations on small trend sections in data. Shape States are identified in sensor data through the orthonormal polynomials method where the polynomial coefficients are used as shape descriptor for the template shape states. A distributed state recognition system has been implemented as an embodiment of the proposed framework and as a tool of verifying the proposed methods. Specific sub-systems of a drilling rig have been used as example systems whose states can be identified. The sub-systems are: Circulation Sub-system, Rotary Sub-system, and Hoisting Sub-system. The verification process of the recognized states is automatically performed and verified against manually classified states from experts. It is proposed to apply the framework and the concept to analyze the drilling rig performance and optimize the drilling process.